How to connect?
The cluster build from a regular (vanilla) PostgreSQL you can use standard connection.
We do recommend use patched PgBouncer per node, so you better to connect using PgBouncer in this case.
The PgBouncer use to adopt XA technique to have a cluster consistency.
Strong or Eventual consistency?
The simple answer: It is configurable per table.
Even more - you can exclude table from a cluster synchronization and make them local.
What about requirements?
PostgreSQL and YTserv must locate on the same linux OS to utilize Unix Domain Socket for faster process communication.
We do recommend use patched PgBouncer, so one more process will be on this OS.
The PgBouncer use to adopt XA technique to have a cluster consistency.
You need one port to listen on PgBouncer or PostgreSQL AND one port for intranode communication.
How to test it?
The testing is very simple and can be found at our github repo .
You can run a cluster simulation as few processes 2-3 per node on same machine.
How to migrate data?
You do not need to migrate data neither into cluster to set up one NOR from cluster into single database to get back from cluster.
The cluster grow-up at runtime is not yet implemented, but we will be working on it. Please support.